ORIGINAL PAPER
Body Odor Quality Predicts Behavioral Attractiveness in Humans
S. Craig Roberts • Alexandra Kralevich • Camille Ferdenzi •
Tamsin K. Saxton • Benedict C. Jones • Lisa M. DeBruine •
Anthony C. Little • Jan Havlicek
Published online: 31 August 2011
 Springer Science+Business Media, LLC 2011
Abstract Growing effort is being made to understand how
different attractive physical traits co-vary within individuals,
partlybecause thismight indicateanunderlying indexofgenetic
quality. In humans, attention has focused on potential markers of
quality such as facial attractiveness, axillary odor quality, the
second-to-fourth digit (2D:4D) ratio and body mass index
(BMI). Here we extend this approach to include visually-
assessed kinesic cues (nonverbal behavior linked to movement)
which are statistically independent of structural physical traits.
The utility of such kinesic cues in mate assessment is controver-
sial, particularly during everyday conversational contexts, as
they could be unreliable and susceptible to deception. However,
we show here that the attractiveness of nonverbal behavior, in
20 male participants, is predicted by perceived quality of their
axillary body odor. This finding indicates covariation between
two desirable traits in different sensory modalities. Depending
on two different rating contexts (either a simple attractiveness
rating or a rating for long-term partners by 10 female raters not
using hormonal contraception), we also found significant rela-
tionships between perceived attractiveness of nonverbal behav-
ior and BMI, and between axillary odor ratings and 2D:4D ratio.
Axillary odor pleasantness was the single attribute that consis-
tently predicted attractiveness of nonverbal behavior. Our results
demonstrate that nonverbal kinesic cues could reliably reveal
mate quality, at least in males, and could corroborate and contrib-
ute to mate assessment based on other physical traits.
Keywords Mate choice  Nonverbal behavior 
Sexual selection  Olfaction  Good genes
Introduction
Many different physical human traits are associated with higher
attractiveness judgments and mate preferences (for reviews, see
Gangestad & Scheyd, 2005; Grammer, Fink, Moller, &
Thornhill, 2003; Rhodes, 2006; Roberts & Little, 2008).
Judgements of facial attractiveness have received most intense
attention in view of the face’s role in human social interactions
(Little & Perrett, 2002; Rhodes, 2006; Roberts, Little et al.,
2005; Thornhill & Gangestad, 1999). Other attributes con-
tributing toattractiveness judgments includemeasuresofbody
dimensions, such as body mass index (BMI) (Maisey, Vale,
Cornelissen, & Tovee, 1999; Pawlowski & Jasienska,2008)
and traits in other sensory modalities, such as voice (Feinberg,
DeBruine, Jones, & Little, 2008; Feinberg et al., 2005; Saxton,
Caryl, & Roberts, 2006; Saxton, DeBruine, Jones, Little, &
Roberts, 2009) and body odor (Havlicek, Dvorakova, Bartos,
S. C. Roberts (&)  A. Kralevich  C. Ferdenzi  T. K. Saxton
School of Biological Sciences, University of Liverpool, Liverpool,
UK
e-mail: craig.roberts@stir.ac.uk
S. C. Roberts  A. C. Little
School of Natural Sciences, University of Stirling, Stirling FK9 4LA,
Scotland, UK
C. Ferdenzi
Swiss Center for Affective Sciences, University of Geneva, Geneva,
Switzerland
T. K. Saxton
Division of Psychology, School of Social and Health Sciences,
University of Abertay Dundee, Dundee, UK
B. C. Jones  L. M. DeBruine
School of Psychology, University of Aberdeen, Aberdeen, UK
J. Havlicek
Department of Anthropology, Charles University, Prague, Czech
Republic
123
Arch Sex Behav (2011) 40:1111–1117
DOI 10.1007/s10508-011-9803-8

& Flegr, 2006; Havlicek, Roberts, & Flegr, 2005; Rikowski &
Grammer, 1999).
These and other sexually selected traits are thought to be
potential indicators of underlying‘‘good genes’’and, as such, are
predicted to co-vary within individuals (Roberts & Little, 2008;
Thornhill &Grammer, 1999). Indeed, intercorrelationsbetween
traits are widely-reported. For example, correlations are seen
between face and body attractiveness (Feinberg et al., 2005;
Saxton, Burriss, Murray, Rowland, & Roberts, 2009), between
facial and body odor attractiveness (Rikowski & Grammer,
1999; Thornhill et al., 2003), and between facial attractiveness
and thesecond-to-fourthdigit ratio (2D:4D,where lowratiosare
related to masculinity) (Ferdenzi, Lemaıˆtre, Leongo´mez, &
Roberts, 2011). Beyond attractiveness, facial masculinity is
related to facial symmetry (Gangestad & Thornhill, 2003;
Little et al., 2008) and is also related to the 2D:4D ratio, where
low ratios are related to masculinity (Burriss, Little, & Nelson,
2007; Fink,Manning, Neave,&Grammer,2004;Fink,Seydel,
Manning, & Kappeler, 2007; Manning, 2002; but see Ferdenzi
et al., 2011).
More recently, evolutionary approaches to human attrac-
tiveness have begun to focus on how judgements are formed
when participants viewstimuli that incorporatemovement. That
is, researchers are investigating attractiveness using judgements
based on so-called dynamic images, including video-recordings
and clips formed from motion-capture devices (e.g., Brown
et al., 2005; Hugill, Fink, & Neave, 2010; Morrison, Gralewski,
Campbell, & Penton-Voak, 2007; Penton-Voak, Allen, Morri-
son, Gralewski, & Campbell, 2007). In contrast with static
images (e.g., photographs, line-drawings), dynamic stimuli con-
tain information that is revealed by movement and includes ges-
tures and other nonverbal behavior (collectively termed kinesic
cues). However, the extent to which such cues might contain reli-
able information about individual quality, in the way thatphysical
traits appear to do, remains somewhat controversial because they
maybefar lessreliable intermsofbiologicalsignaling(Grammer,
Honda,Juette,&Schmitt,1999;Roberts,2008).Forexample, they
may be temporally variable, context-dependent, culturally influ-
enced, and potentially susceptible to manipulation and deception
(Roberts, 2008). While this must be true to some extent, some
evidence exists to suggest that such cues do reliably influence the
perception and judgments of others. For example, in an observa-
tional study, the frequency of several specific behavioral patterns
exhibited by men in a bar predicted their subsequent success in
establishingconversational relationshipswithwomen (Renninger,
Wade, & Grammer, 2004).
If such outcomes are meaningful in revealing genetic quality
with a reasonable degree of reliability, then at least some of this
information is also predicted to correlate with relevant physical
traits used in mate choice (Hugill et al., 2010; Roberts, 2008).
Thereexistssomeevidencefor this, asBrownetal. (2005)showed
that variation in attractiveness of dancing ability, a dynamic dis-
play used cross-culturally in human courtship, correlates with
body symmetry. Furthermore, men with low (i.e., masculine)
2D:4D ratios were judged to dance more attractively than men
with more feminized ratios (Fink et al., 2007). While these studies
provide importantpreliminaryevidencefora linkbetweenkinesic
information and physical indicator traits, the ability to dance in an
attractivemanner is likelylinkedtoameasureofco-ordinationand
athleticismwhichisexhibitedonlyinparticularcircumstancesand
relatively rarely compared to normal conversational contexts in
day-to-day behavior. Here, we aimed to extend such research by
investigating correlations between a set of phenotypic indicator
traits and kinesic information within a more common scenario–
simply observing someone talking.
A major obstacle for efforts to understand how behavioral
cues affect quality perception by potential mates is that dynamic
stimuli contain relevant information about both behavior and
other influential physical attributes. It is, therefore, difficult to
disentangle the effects of behavioral cues and those of the other
traits on perceptions of judges. Thus, it is no surprise that ratings
of human faces in static and dynamic form are intercorrelated
(Riggio, Widaman, Tucker, & Salinas, 1991; Roberts, Saxton
et al., 2009), even though some studies have suggested that the
levelof thiscorrelation is lower thanmightbeexpected (Lander,
2008; Penton-Voak & Chang, 2008; Rubenstein, 2005). One
solution to this problem is to employ new digital technologies to
capture movement independently of shape or other information,
by using wire-frame images or standardized models (Brown
et al., 2005; Morrison et al., 2007). However, such approaches
are not without their own problems. For example, generation of
wire-frame images is best-suited to studies of gait (or dance)
where thescaleofmovement is relativelyexpansive; itperforms
relatively poorly at capturing finer-scale movement, especially
facial expressions. It also often requires markers to be attached
to the body and time for spatial calibration (potentially altering
or inhibiting normal behavior), and requires expensive motion-
capture equipment. A more accessible solution is to statistically
partial out the influence of structural physical features on
judgments by calculating residuals from static-dynamic image
regression, providing researchers with a standardized measure
of attractiveness based on movement cues alone (Roberts, Little
et al., 2009; Roberts, Saxton et al., 2009).
In thisstudy,weused thisapproachfor thefirst timeto investi-
gate the intercorrelations between attractive dynamic cues and
putative indicator traits: axillary odor, 2D:4D ratio, and BMI. In
addition, we examined the relationships between odor and either
2D:4DratioorBMI,neitherofwhichhavebeenpreviouslyinves-
tigated. We photographed and video-recorded young men, col-
lected samples of their body odor, and measured the relevant traits.
We then asked a group of female raters to judge the photos, videos,
and odors.
1112 Arch Sex Behav (2011) 40:1111–1117
123